I wonder how machine learning, particularly deep neural net which is considered the state-of-the-art among many data scientists, is used among economists, or other experts who is in in the similar field to predict economic situations, such as bank analysts, etc...

I know in econometrics people heavily use statistics. But it is focused on statistics, not on machine learning, let alone deep learning.

How prevalent is it? What kind of methods/algorithms/models, if any, are frequently used for what kind of purposes among what kind of economists? Where can I see their work (e.g. blog, paper, etc...)

  • 1
    $\begingroup$ "How prevalent" is hard to measure. I guess that is why two people have already voted to close this question. Perhaps you would be better off asking for examples of famous econ papers that use machine learning. $\endgroup$
    – Giskard
    Aug 11, 2017 at 7:17
  • $\begingroup$ It's heavily uesd in financial and economic forecasting. But if a method works, it won't be published. If it's published, it doesn't work. $\endgroup$
    – 410 gone
    Aug 11, 2017 at 7:24
  • 1
    $\begingroup$ If it's published, it will still possibly work if the primary author is an assistant professor. $\endgroup$
    – High GPA
    Aug 11, 2017 at 11:14
  • 1
    $\begingroup$ First, you need to pull up the economist jokes. In fact, ML is just all about garbage in garbage out. Garbage economic assumption will return garbage output. $\endgroup$
    – mootmoot
    Aug 11, 2017 at 15:22
  • $\begingroup$ @denesp Sorry if that is too broad but I don't see the closed votes casted... $\endgroup$
    – Blaszard
    Aug 13, 2017 at 10:01

2 Answers 2


I will attempt to answer the first (How prevalent is it?) and last question (Where can I see their work?).

There was a recent post in The Economist about the emergence of machine learning into standard economic papers. They produced a nice graph, reproduced below:

enter image description here

Albeit machine learning is confounded with big data, it shows a rise in their use since 2014. Surely, this is a very narrow selection of the literature (NBER working papers, and only abstracts), but is a start.

Regarding where you can find papers using machine learning, you can use "keyword" search here (26,665 documents are found), or here (337 documents are found). The latter allows you to select dates, which you could use to build your own indicator of machine learning related publications per year.

Finally, related to blogs, I am not aware of a dedicated ML/DL economist's blog. However, here there is a very useful blog aggregator, which has indexed most of the "serious" blogs available. The full list of blog posts indexed is here. A search for "machine" or "deep" does not yield a result. However, there are some hits for "econometric" and "statistics" which might help you.


As previously stated by EnergyNumbers, ML is very often used in financial and economics forecasting, just like OLS. The deep learning is less useful in economics academia because you do not know the cause and it will be hard to publish a well-explained paper based on that.

What kind of methods/algorithms/models, if any, are frequently used for what kind of purposes among what kind of economists? Where can I see their work (e.g. blog, paper, etc...)

You could just use goole scholar and search for "financial time series forecasting machine learning". GARCH, SVM, and ARIMA are often used. You could get started here:

Kim, Kyoung-jae. "Financial time series forecasting using support vector machines." Neurocomputing 55.1 (2003): 307-319.

Hsu, Ming-Wei, et al. "Bridging the divide in financial market forecasting: machine learners vs. financial economists." Expert Systems with Applications 61 (2016): 215-234.

Shynkevich, Yauheniya, et al. "Forecasting stock price directional movements using technical indicators: investigating window size effects on one-step-ahead forecasting." Computational Intelligence for Financial Engineering & Economics (CIFEr), 2104 IEEE Conference on. IEEE, 2014.

Tan, Elliot. "Forecasting Foreign Exchange Rates in Response to Federal Reserve Communication: A Machine Learning Approach." (2016).

  • 2
    $\begingroup$ GARCH and ARIMA hardly are machine learning approaches; rather, they are traditional tools in financial econometrics. (You could of course go as far as to call regression analysis a machine learning techique; but then the distinction between traditional econometric methods and machine learning would be blurry.) SVM, on the other hand, is a "good" example of machine learning. $\endgroup$ Aug 11, 2017 at 18:39
  • $\begingroup$ Thanks but then why is SVM used while DNN not? Seems to me that SVM is also pretty hard to detect the cause. $\endgroup$
    – Blaszard
    Aug 13, 2017 at 9:57
  • $\begingroup$ @Blaszard DNN is used, but less than SVM, because of the number of parameters (over-fitting). Moreover, using SVM you need less experiments and to spend less time play-around. There are also other minor issues such as global optimum, etc. SVM is definitely more interpretable than DNN. $\endgroup$
    – High GPA
    Aug 15, 2017 at 15:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.